粮食作物病虫害光谱监测研究进展

Research progress in spectral monitoring of diseases and pests in cereal crops

关键词:  作物病虫害  遥感监测  病虫害识别  严重度估算  病虫害制图

Key words:  crop pests and diseases  remote sensing monitoring  pests and diseases identification  severity quantification  pests and diseases mapping

摘要:

【目的】作物病虫害遥感监测是精准防控农田生物胁迫、保障国家粮食安全的关键技术。现有研究在特征筛选和建模方法上的进展显著提高了病虫害监测精度，但与解决生产需求之间还存在差距，主要表现为模型精度有限且泛化能力不强。【方法】本文梳理了水稻、小麦、玉米三大粮食作物病虫害遥感监测研究最新成果，从光谱监测机理、遥感数据选择、建模制图方法、生产与育种应用等角度分析了现有研究的发展脉络与技术瓶颈。【结果】现有研究基于不同作物病虫害机理筛选特异性多源特征组合，构建了从经典机器学习向机理引导的深度学习发展的算法框架，各类病虫害监测效果显著提升。但受限于模型时空迁移能力及诊断决策方法的缺失，现有技术面对动态场景难以形成智能决策，与满足实际大田病虫害防控需求间存在差距。【结论】为打破作物病虫害光谱监测理论构建与田间应用间的壁垒，未来研究需进一步深入机理解析、数据扩展与模型优化，构建天空地多尺度智能监测与决策体系，从而在农田绿色智能防控中发挥实质性作用，为发展农业新质生产力提供坚实技术支撑。

Abstract:

[Purpose] Remote sensing monitoring of crop pests and diseases is a critical technology for precisely controlling farmland biotic stress and ensuring national food security. Substantial advances in feature selection and modeling methodologies have considerably improved monitoring accuracy. However, a gap remains between current research outcomes and practical production requirements, primarily manifested in limited model accuracy and insufficient generalization capability. [Method] This paper reviews the latest research achievements in remote sensing monitoring of pests and diseases for three major food crops: rice, wheat, and maize. The evolutionary trajectory and technical bottlenecks of existing studies are systematically analyzed from the perspectives of spectral monitoring mechanisms, remote sensing data selection, modeling and mapping methodologies, as well as production and breeding applications. [Result] Existing studies have selected specific multi-source feature combinations based on different crop pest and disease mechanisms, and constructed an algorithmic framework evolving from classical machine learning toward mechanism-guided deep learning, leading to a significant improvement in the monitoring efficacy of various pests and diseases. Nevertheless, constrained by the limited spatiotemporal transferability of models and the absence of diagnostic decision-making frameworks, existing technologies struggle to make intelligent decisions in dynamic scenarios, leaving a gap in meeting the practical demands of field pests and diseases prevention and management. [Conclusion] To break down the barriers between spectral monitoring theory and field application for crop pests and diseases, future research must deepen mechanism analysis, data expansion, and model optimization to construct a space-air-ground multiscale intelligent monitoring and decision-making system. This will enable these technologies to play a substantive role in green and intelligent farmland prevention and management, thereby providing solid technical support for developing new quality productive forces in agriculture.